Fig. 4.

ROC curves of different machine learning models of training set: (A) PLS; (B) RF; (C) GBM; (D) SVM. ROC curves of different machine learning models of test set: (E) PLS; (F) RF; (G) GBM; (H) SVM. (I) Computational times of different machine learning models. (J) Predictive performance of six single-metabolite and four multiple-metabolite-based machine learning models in the test set. PLS: partial least-square; RF: random forest; GBM: gradient boosting machine; SVM: support vector machine.